The first half of this week was spent in trying to debug the cause for the “divide by zero” issue. Last week, I had come to a conclusion that the problem was with the the graph that was generated from the ontologies. However, when I queried the graph with other sparql queries, correct results were returned. Also, when the graph was queried after splitting existing sparql statements into smaller segments, it returned results that were verified as correct after manual evaluation of the triples. Since the other members of the team were running the code without any problem, we assumed that there was some problem either with my machine or the way in which I had setup the project. Later, we found out that there was some issues in the URLs that were being used and since other members were using the old version of the code, there had not encountered the problem. This issue was finally resolved and I was successfully able to run the code. Currently, I have been making changes in the method for comparing the contents for similarity and have been evaluating its effect.
In the second half of the week, I started setting up AgreementMakerLight (AML) tool for ontology matching. I explored the existing matching algorithms present in AML. I also became familiar with the basic algorithm that was used by one of my team members for matching ontologies. I plan to add new modules that can compare abbreviations to their expanded form and can identify units and their corresponding physical quantities to the existing algorithm. After thoroughly exploring the dataset, I believe that these components can help in improving accuracy. Hence, next week I will focus on implementing these components and will simultaneously work on the Linkipedia code too.